Advanced fetal cardiovascular diagnostic imaging by model-constrained undersampled MRI and deep learning

Fetal cardiovascular MRI has been developed at Lund University over the last eight years. It is a state-of-the-art research tool aiming to diagnose disease before the patient is born. In-utero quantitative imaging of the cardiovascular system presents unique challenges such as 1) difficulty to synchronize the MRI data to the cardiac motion due to a weak and contaminated fetal electrocardiogram along with high heart rates, 2) fetal motion, and 3) a small imaging target.

We have developed an MRI platform that provides 3D movies of the beating fetal heart. We aim to implement novel acquisition and post-processing techniques to make this tool clinically useful for interrogating fetal cardiovascular pathophysiology in congenital cardiac malformation and to uncover the mechanisms and pathophysiology in dysfunctional placenta and compromised utero-placental blood flow in relation to its consequences on fetal cardiovascular disease, which is vital for understanding, prevention and potential cure.

We will design new MRI data acquisition and reconstruction algorithms based on radial acquisitions, undersampled datasets, deep learning and simulated datasets for training. The goal is to compensate for fetal motion to improve image quality and reduce exam times. We will develop a fast reconstruction engine based deep learning to generate images in minutes.

This project could impact decision-making before birth worldwide and have a direct impact on delivery planning and care after birth.